Part of endosymbiotic theory is pretty well accepted now (I remember being told in school that mitochondria may have come from prokaryotic cells), but it's a bit optimistic to think that symbiosis causes all speciation. It's very conceivable that most speciation is caused through many generations of reproduction, environmental stressors and separation.
But one can imagine a (perhaps more boring) applet where the computer makes the choice for you. Maybe the computer always selects the biggest critter to reproduce. Then, even though there is no intelligent overseer, the critters will tend to get larger and larger. Or one can imagine a more complicated selection process involving a physics simulation where a bunch of copies of the critter are dropped on top of one another in a pile. The tallest pile is selected to reproduce. Over time, one would tend to see critters that were good at stacking on top of one another, maybe they would be rectangular like bricks, or maybe they would have lots of spines so they could stick together like burrs. Or maybe evolution would hit upon something else. But there is no need for the selection process to be intelligent.
Of course, in those examples, you might say that whoever wrote the objective, like "grow big", or "stack in tall piles" is still directing the process from behind the computer program. That's why in the natural world, we don't have objectives like that. Instead, the only "objective" in the natural world is "survive and reproduce", since that is tautologically what determines reproductive success. There is a freely available software application called "biogenesis" that shows evolution using that objective function in a world with greatly simplified physics, if you are curious.
There are certainly plenty of examples of evolution being driven by the natural environment, even if the books tend to focus on things like evolutionary arms races, symbiosis, and sexual selection. Here is one: An organism that can convert a molecule of glucose to more molecules of ATP than its neighbours can will be able to use its food more efficiently than them. This will give it an edge in survival and reproduction, since it needs less food to do the same amount of stuff. So the genes that enable this more efficient metabolism will tend to spread throughout the population, because the organisms that bear them will be less likely to die of starvation.
So, even if most of the modifications that viruses make are deleterious, as you say, if a virus comes along that confers some advantage, it is much more likely that the resulting change will be in play in the descendants. A beneficial mutation on the other hand would have to somehow be happening in some sort of Genghis Khan creature with scores of offspring to have the same reach as a beneficial virus. So you can see the probable relative importance of these two bases of variation.